I am new to python and am trying to some simple classification on raster image.
Basically, I am reading a TIF image as a 2D array and do some calculating and manipulation on it. For classification part, I am trying to create 3 empty arrays for land, water, and clouds. These classes will be assigned a value of 1 under multiple conditions, and eventually assigning these classes as landclass=1, waterclass=2, cloudclass=3 respectively.
apparently I can assign all values in an array to 1 under one condition
like this:
crop = gdal.Open(crop,GA_ReadOnly)
crop = crop.ReadAsArray()
rows,cols = crop.shape
mode = int(stats.mode(crop, axis=None)[0])
water = np.empty(shape(row,cols),dtype=float32)
land = water
clouds = water
than I have something like this (output):
>>> land
array([[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
...,
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.]], dtype=float32)
>>> land[water==0]=1
>>> land
array([[ 0., 0., 0., ..., 1., 1., 1.],
[ 0., 0., 0., ..., 1., 1., 1.],
[ 0., 0., 0., ..., 1., 1., 1.],
...,
[ 1., 1., 1., ..., 0., 0., 0.],
[ 1., 1., 1., ..., 0., 0., 0.],
[ 1., 1., 1., ..., 0., 0., 0.]], dtype=float32)
>>> land[crop>mode]=1
>>> land
array([[ 0., 0., 0., ..., 1., 1., 1.],
[ 0., 0., 0., ..., 1., 1., 1.],
[ 0., 0., 0., ..., 1., 1., 1.],
...,
[ 1., 1., 1., ..., 0., 0., 0.],
[ 1., 1., 1., ..., 0., 0., 0.],
[ 1., 1., 1., ..., 0., 0., 0.]], dtype=float32)
But how can I have the values in "land" equal to 1 under a couple of conditions without altering the shape of the array?
I tried to do this
land[water==0,crop>mode]=1
and I got ValueError. And I tried this
land[water==0 and crop>mode]=1
and python asks me to use a.all() or a.all()....
For only one condition, the result is exactly what I want, and I have to do it in order to get the result. eg (this is what I have in my actual code):
water[band6 < b6_threshold]=1
water[band7 < b7_threshold_1]=1
water[band6 > b6_threshold]=1
water[band7 < b7_threshold_2]=1
land[band6 > b6_threshold]=1
land[band7 > b7_threshold_2]=1
land[clouds == 1]=1
land[water == 1]=1
land[b1b4 < 0.5]=1
land[band3 < 0.1)]=1
clouds[land == 0]=1
clouds[water == 0]=1
clouds[band6 < (b6_mode-4)]=1
I found this is a bit confusing and I would like to combine all conditions within one statement... Any suggestion on that?
Thank you very much!
You can multiply the boolean arrays for something like "and":
>>> import numpy as np
>>> a = np.array([1,2,3,4])
>>> a[(a > 1) * (a < 3)] = 99
>>> a
array([ 1, 99, 3, 4])
And you can add them for something like "or":
>>> a[(a > 1) + (a < 3)] = 123
>>> a
array([123, 123, 123, 123])
Alternatively, if you prefer to think of boolean logic rather than True and False being 0 and 1, you can also use the operators & and | to the same effect.
Related
I have arrays of size A (1, 1, 59) and B (1, 95, 59). I want to concatenate the arrays. The size of the array should be (1, 96, 59).
np.concatenate((A, B),axis =0)
Doesn't work. The error is ValueError: all the input array dimensions except for the concatenation axis must match exactly
The axis is incorrect:
>>> import numpy as np
>>> A = np.ones((1,1,59))
>>> B = np.zeros((1,56,59))
>>> np.concatenate((A, B), axis=1)
array([[[ 1., 1., 1., ..., 1., 1., 1.],
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
...,
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.]]])
>>> _.shape
(1, 57, 59)
I have an array of zeros
arr = np.zeros([5,5])
array([[ 0., 0., 0., 0., 0.],
[ 0., 0., 0., 0., 0.],
[ 0., 0., 0., 0., 0.],
[ 0., 0., 0., 0., 0.],
[ 0., 0., 0., 0., 0.]])
I want to assign values based on index so I did this .
out = np.array([[nan,2.,4.,1.,1.],[nan,3.,4.,4.,4.]])
arr[out[0].astype(int),np.arange(len(out[0]))] = 1
arr[out[1].astype(int),np.arange(len(out[1]))] = 1
Assignment works fine if there is 0 instead of nan.
How can I skip assignment in case of nan? and Is it possible to assign values at once from a multidimensional index array rather than using for loop ?
Mask it -
mask = ~np.isnan(out)
arr[out[0,mask[0]].astype(int),np.flatnonzero(mask[0])] = 1
arr[out[1,mask[1]].astype(int),np.flatnonzero(mask[1])] = 1
Sample run -
In [171]: out
Out[171]:
array([[ nan, 2., 4., 1., 1.],
[ nan, 3., 4., 4., 4.]])
In [172]: mask = ~np.isnan(out)
...: arr[out[0,mask[0]].astype(int),np.flatnonzero(mask[0])] = 1
...: arr[out[1,mask[1]].astype(int),np.flatnonzero(mask[1])] = 1
...:
In [173]: arr
Out[173]:
array([[ 0., 0., 0., 0., 0.],
[ 0., 0., 0., 1., 1.],
[ 0., 1., 0., 0., 0.],
[ 0., 1., 0., 0., 0.],
[ 0., 0., 1., 1., 1.]])
Alternative, replace the flatnonzero calls with range-masking -
r = np.arange(arr.shape[1])
arr[out[0,mask[0]].astype(int),r[mask[0]]] = 1
arr[out[1,mask[1]].astype(int),r[mask[1]]] = 1
If you are working with a lot many rows than just 2 and you want to assign them in a vectorized manner, here's one method, using linear-indexing -
n = arr.shape[1]
linear_idx = (out*n + np.arange(n))
np.put(arr, linear_idx[~np.isnan(linear_idx)].astype(int), 1)
Given a numpy array of zeros, say
arr = np.zeros((5, 5))
and an array of indices that represent vertices of a polygon, say
verts = np.array([[0, 2], [2, 0], [2, 4]])
1) What is the elegant way of doing
for v in verts:
arr[v[0], v[1]] = 1
such that the resulting array is
In [108]: arr
Out[108]:
array([[ 0., 0., 1., 0., 0.],
[ 0., 0., 0., 0., 0.],
[ 1., 0., 0., 0., 1.],
[ 0., 0., 0., 0., 0.],
[ 0., 0., 0., 0., 0.]])
2) How can I fill the array with ones such that the output array is
In [158]: arr
Out[158]:
array([[ 0., 0., 1., 0., 0.],
[ 0., 1., 1., 1., 0.],
[ 1., 1., 1., 1., 1.],
[ 0., 0., 0., 0., 0.],
[ 0., 0., 0., 0., 0.]])
To answer the first part of your question: arr[tuple(verts.T)] = 1
verts.T transposes your indices to a (2, n) array, where the two rows correspond to the row and column dimensions of arr. These are then unpacked into a tuple of (row_indices, col_indices), which we then use to index into arr.
We could write this a bit more verbosely as:
row_indices = verts[:, 0]
col_indices = verts[:, 1]
arr[row_indices, col_indices] = 1
For the second part, one method that will work for arbitrary polygons would be to use matplotlib.Path.contains_points, as described here:
from matplotlib.path import Path
points = np.indices(arr.shape).reshape(2, -1).T
path = Path(verts)
mask = path.contains_points(points, radius=1e-9)
mask = mask.reshape(arr.shape).astype(arr.dtype)
print(repr(mask))
# array([[ 0., 0., 1., 0., 0.],
# [ 0., 1., 1., 1., 0.],
# [ 1., 1., 1., 1., 1.],
# [ 0., 0., 0., 0., 0.],
# [ 0., 0., 0., 0., 0.]])
This question already has answers here:
How do I print the full NumPy array, without truncation?
(22 answers)
Closed 9 years ago.
I am working with image processing in python and I want to output a variable, right now the variable b is a numpy array with shape (200,200). When I do print b all I see is:
array([[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
...,
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.]])
How do I print out the full contents of this array, write it to a file or something simple so I can just look at the contents in full?
Of course, you can change the print threshold of the array as answered elsewhere with:
np.set_printoptions(threshold=np.nan)
But depending on what you're trying to look at, there's probably a better way to do that. For example, if your array truly is mostly zeros as you've shown, and you want to check whether it has values that are nonzero, you might look at things like:
import numpy as np
import matplotlib.pyplot as plt
In [1]: a = np.zeros((100,100))
In [2]: a
Out[2]:
array([[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
...,
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.]])
Change some values:
In [3]: a[4:19,5:20] = 1
And it still looks the same:
In [4]: a
Out[4]:
array([[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
...,
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.]])
Check some things that don't require manually looking at all values:
In [5]: a.sum()
Out[5]: 225.0
In [6]: a.mean()
Out[6]: 0.022499999999999999
Or plot it:
In [7]: plt.imshow(a)
Out[7]: <matplotlib.image.AxesImage at 0x1043d4b50>
Or save to a file:
In [11]: np.savetxt('file.txt', a)
to_print = "\n".join([", ".join(row) for row in b])
print (to_print) #console
f = open("path-to-file", "w")
f.write(to_print) #to file
In case it's numpy array: Print the full numpy array
I am trying to use reduce() function to create a function hstack() which horizontally stacks multiple arrays. As a simple example, lets say
>>>>M=eye((4))
>>>>M
array([[ 1., 0., 0., 0.],
[ 0., 1., 0., 0.],
[ 0., 0., 1., 0.],
[ 0., 0., 0., 1.]])
>>>>hstack([M,M])
array([[ 1., 0., 0., 0., 1., 0., 0., 0.],
[ 0., 1., 0., 0., 0., 1., 0., 0.],
[ 0., 0., 1., 0., 0., 0., 1., 0.],
[ 0., 0., 0., 1., 0., 0., 0., 1.]])
This works as I want. Now I define
>>>> hstackm = lambda *args: reduce(hstack, args)
And try to do the hstack() from the previous case
>>>>hstackm([M,M])
[array([[ 1., 0., 0., 0.],
[ 0., 1., 0., 0.],
[ 0., 0., 1., 0.],
[ 0., 0., 0., 1.]]),
array([[ 1., 0., 0., 0.],
[ 0., 1., 0., 0.],
[ 0., 0., 1., 0.],
[ 0., 0., 0., 1.]])]
Which is incorrect. How do I define hstackm() to obtain a proper output?
My final objective will be to create a hstackm() function to stack SPARSE matrices if it is possible. Something like,
hstackm = lambda *args: reduce(sparse.hstack, args).
The _*args_ would be csr or _lil_matrix_
thank you
In [16]: hstackm = lambda args: reduce(lambda x,y:hstack((x,y)), args)
In [17]: hstackm([M,M])
Out[17]:
array([[ 1., 0., 0., 0., 1., 0., 0., 0.],
[ 0., 1., 0., 0., 0., 1., 0., 0.],
[ 0., 0., 1., 0., 0., 0., 1., 0.],
[ 0., 0., 0., 1., 0., 0., 0., 1.]])
Your function hstack takes one parameter, a list of matrices. reduce() calls it with two parameters instead, each a matrix.
Change your hstack method to accept an arbitrary number of arguments instead:
def hstack(*matrices):
....
instead of hstack(matrices), then call it as hstack(M, M).